[reportlab-users] Error reporting: please say what image is failing

[Robin Becker]

On 18/03/2010 10:38, Andy Robinson wrote:

> On 18 March 2010 09:55, Robin Becker<robin at reportlab.com>  wrote:

>> we can if it's given as a string, and perhaps, as a file like object. Maybe

>> Roberto can give some indication of where the image comes from and perhaps a

>> sample.

>

>

> One more idea:  if people are passing in file-like objects, perhaps we

> could allow an optional argument (audit_trail="foo.jpg"?)  to the

> image constructor or function.  If an error occurs and we were not

> given a filename or URL, we could add that argument to the args for

> the exception.


the place for this information is on the filelike object. I believe that is 
intended for files that are opened by path eg



>>>> f=file('/tmp/aaa','r')

>>>> f.name

> '/tmp/aaa'

>>>>


so logically we should try for obj.name as that's the common case. For the 
standard StringIO case nothing prevents use of a name on the object.


>>>> import StringIO

>>>> s=StringIO.StringIO(buf='aaaa')

>>>> s.name='aaaa'

>>>>


but that won't work for cStringIO and may not for sockets. We're already trying 
to handle the case of PIL images being passed in.




>

> Examples:  some systems have big databases of images or pull them from

> a database connection, so will be passing us file-like objects.  The

> programmer can then separately add an identifier.

>


the danger with names at the point of call is that they often end up as 
completely generic names from a template location. I think we should allow an 
ImageReader identifier, but also report any name decided from the file pointer 
since that will be more specific.


> Roberto, how should rst2pdf behave if an image is missing?  Should it

> fail on first error, or does it have any modes to continue-but-warn,

> or put dummy image placeholders in the PDF?      You probably have

> issues we have had in the past with RML...

>


Looking in the ImageReader code there is a specific exception being made for PIL 
1.1.7 because that did cause a problem with split ie

	if mode=='RGBA':
		if Image.VERSION.startswith('1.1.7'): im.load()
		self._dataA = ImageReader(im.split()[3])
		im = im.convert('RGB')
		self.mode = 'RGB'

can we find some details on the PIL version etc etc.
.......
-- 
Robin Becker